Elderly patients, who comprise a growing majority of patients undergoing inpatient surgery, experience markedly higher rates of morbidity and mortality than their younger counterparts. In light of widespread recognition that surgical quality varies widely, there is growing demand for providing patients with information about hospital performance. Unfortunately, reliable quality indicators are lacking for most procedures. Procedure volume does not reliably predict mortality of individual hospitals for most procedures. Process measures, the basis of most current pay for performance plans, generally relate to secondary outcomes. And, finally, direct outcome measures (e.g., risk-adjusted mortality) are too "noisy" to reflect performance at most hospitals. For these reasons, we propose a novel, integrative measure of quality that makes optimal use of all available information related to hospital mortality with a specific procedure. Our application has 2 specific aims: I. To develop integrative measures of surgical performance. Using data from the national Medicare population (1998-2001), we will develop integrative performance measures for each of 10 common or high-risk inpatient surgical procedures. Model inputs will include measures from different quality domains, including structural variables (e.g., procedure volume), process of care, and patient outcomes (both for the procedure of interest and related operations). Model outputs will be hospital- and procedure-specific estimates of "true" mortality that optimally filter out statistical noise. II. To assess how well integrative measures predict future performance. Given how patients hope to use them, surgical quality indicators are best judged by whether historical measures predict future performance. In this context, we will again use Medicare data to assess how well our integrative measures (from 1998-2001 data) predict hospital mortality rates in 2002-2003, relative to individual quality indicators (e.g., historical mortality or volume alone). [unreadable] [unreadable] [unreadable]